Neural Network Fault Diagnosis System for a Diesel-Electric Locomotive's Closed Loop Excitation Control System

Abstract

In closed loop control systems fault isolation becomes extremely difficult in the case of feedbacks being oscillatory due to corrupted signals or malfunctions in actuators. This paper investigates and highlights the development of an off-line fault detection and isolation system for the isolation of faults, which cause oscillatory conditions on a General Electric (GE) Diesel-Electric Locomotive's excitation control system. The paper illustrates the use of artificial neural networks as a replacement to classical analytical models used for residual generation. The artificial neural network model's design is based on model-based dedicated observer theory to isolate sensor, as well as component faults, where observer theory is utilised to effectively select input-output data configurations for detection of sensor and component faults causing oscillations. Residual Evaluation is done with the use of a moving average filter incorporated with the simple thresholding technique. The results indicated 100% accuracy for the detection and isolation of the component or sensor responsible for causing excessive oscillation in the excitation control system.